Materialization scheduler Phase 2: CLI + full docs pass (follow-up to #888)

[girishjeswani]

Phase-2 follow-ups to #888, tracked separately so that PR stays focused on the standalone scheduler + config-parity core. Sequenced after it lands.

CLI

Add publisher CLI coverage for the materialization surface (the CLI already has a generated client to build on):

• schedule view / set / clear — read and edit a package's materialization.schedule, sharing the same publish-gate validation the server enforces.
• materializations list — per-package and environment-scoped, with the trigger (ON_DEMAND / SCHEDULER) column.
• Trigger a run on demand.
• Inspect a run's manifest / status.

Agent skills (depends on the CLI)

Cost/performance optimization skills that lean on the CLI + publisher APIs — what's persisted vs not, run timings and sources built/reused from the materialization history — to recommend persist candidates, schedule/scope changes, or removals. Recommendations-only.

Full docs pass (depends on the above)

Once the CLI and local devex are in place:

• Malloy docs site: getting-started, config/env-var reference, the publish-gate rules stated as rules, fire semantics.
• Repo README/docs: operator depth, the environment-scoped view, and an honest list of the standalone-vs-hosted divergences (e.g. tables-only local fires vs the two-phase hosted job).

Docs come last, since the content depends on how the CLI and any remaining gap-closers land.